Author Notes: Covering Three 2020 LHP Draftees
Link to article: https://baseballcloud.blog/2020/06/19/highlighting-bcteam-draftees-jared-shuster-burl-carraway-and-sam-weatherly/
I have never been much of a college or minor leagues guy. I care about the numbers produced in the major leagues and that is it. So when I got the assignment to cover three draftees, it lead to a bit of a scramble. What insight could a newbie college analyst provide to a reader?
Having ample experience with trackman data through Baseball Savant, I decided to go with the angle of MLB comparisons. Baseball Cloud has a HEAFTY database with trackman and flightscope data dating back to 2018. So while there may not be the same information on current major leaguers on their time in college, direct comparisons could be made to the pros today.
For the release point comparisons, I took pitchers who threw 500+ pitches and minimized the following equation for each college pitcher:
This was basically the formula I used for the pitch similarities as well; adding sum the squared differences of each component and finding the lowest value. For the pitch similarities, I took all pitch types thrown 50+ times to compare with than bucketed the into options of fastballs (four seam, two seam, sinkers), changeups, and breaking balls (sliders and curveballs).
Two key decisions I made was keeping it to lefties and not adjusting factors. What I mean by that is rather than flipping righty movement profiles to increase options to compare to, I stuck to left handed MLB pitchers. I did this so when analyzing how the pitch is used by the MLB pitcher, there would be no adjusting for the MLB pitcher being a righty but the college pitcher being a lefty.
Regarding the adjusting factors, I could have divided the differences by the standard deviation of the factor before squaring. This would have made sense as an additional 0.1 MPH is more significant than an additional 0.1 inches of movement. The main reason I did not do this is because putting sinkers in with four seamers and sliders in with curveballs produced a large standard deviation in vertical movement. So Burl Carraway’s curveball had a particularly bad comparison with Josh Osich’s slider. The vertical movement difference was roughly ten inches and the horizontal movement difference was substantial as well. One casualty was Burl’s fastball being compared to Aroldis Chapman’s (large speed gap) instead of Blake Snell’s, but it was a small price to pay relative to the Osich comparison.
Speaking of Burl Carraway’s fastball, its movement was truly absurd. I emphasize in the article its vertical movement, but plugging all his fastball averages into Alan Nathan’s trajectory calculator makes his pitch movement physically impossible at his spin rate. The possible movement at 100% spin efficiency is only 0.4 inches off from his recorded movement (and that’s without playing around with barometric conditions), but it does raise an eyebrow. Chances are that it is due to trackman/flightscope misreads, but it may be a Rich-Hill-type scenario.
Hill’s curveball is highly efficienct and he varies the speeds he throws it. This results in an average curveball spin efficency of 100%. This can theoretically only be possible with a perfect 12–6 curveball or 6–12 fastball, but Hill’s curveball does have a significant component of horizontal sidespin so I imagine its the averaging of pitch components that make them be physically impossible together. Nonetheless, it indicates Burl has an incredibly high spin efficiency fastball.
One more thing that tugged at my sanity was Baseball Savant’s statcast search not making it easy to find each player’s average extension. I could have it output both the pitcher’s average horizontal and vertical releases, but not the extension. I was able to emphasize Burl Carraway’s extension absurdity by specifying extension through the “metric range” option on statcast search, but it was frustrating.
In general, I do not know why there has not been more research on release points. There has been a good amount on repeating release points, both pitch to pitch for tunneling and game to game for command, but not that much player to player. I think analyzing the percentile of the release point components could provide insight into how various players go about making their deliveries more deceptive. That might make for a solid article in the future… guess we will wait and see on that one.
Unused graphic:
